Robust Revenue Maximization Under Minimal Statistical Information

TL;DR

This paper develops simple, robust auction mechanisms for multi-item revenue maximization with minimal prior knowledge, achieving near-optimal guarantees based on the variability of item valuations.

Contribution

It introduces mechanisms that guarantee revenue approximations based on the coefficient of variation, with tight bounds and improvements under independence and identical distribution assumptions.

Findings

Separate price lotteries achieve an $O( ext{log } r)$ approximation.

Deterministic mechanisms have an $O(r^2)$ approximation.

Bundling improves guarantees under independence.

Abstract

We study the problem of multi-dimensional revenue maximization when selling $m$ items to a buyer that has additive valuations for them, drawn from a (possibly correlated) prior distribution. Unlike traditional Bayesian auction design, we assume that the seller has a very restricted knowledge of this prior: they only know the mean $\mu_j$ and an upper bound $\sigma_j$ on the standard deviation of each item's marginal distribution. Our goal is to design mechanisms that achieve good revenue against an ideal optimal auction that has full knowledge of the distribution in advance. Informally, our main contribution is a tight quantification of the interplay between the dispersity of the priors and the aforementioned robust approximation ratio. Furthermore, this can be achieved by very simple selling mechanisms. More precisely, we show that selling the items via separate price lotteries achieves an $O(\log r)$ approximation ratio where $r=\max_j(\sigma_j/\mu_j)$ is the maximum coefficient of variation across the items. To prove the result, we leverage a price lottery for the single-item case. If forced to restrict ourselves to deterministic mechanisms, this guarantee degrades to $O(r^2)$. Assuming independence of the item valuations, these ratios can be further improved by pricing the full bundle. For the case of identical means and variances, in particular, we get a guarantee of $O(\log(r/m))$ which converges to optimality as the number of items grows large. We demonstrate the optimality of the above mechanisms by providing matching lower bounds. Our tight analysis for the single-item deterministic case resolves an open gap from the work of Azar and Micali [ITCS'13]. As a by-product, we also show how one can directly use our upper bounds to improve and extend previous results related to the parametric auctions of Azar et al. [SODA'13].